Collact.Me: Conceptual Framework for Extracting Domain-Specific Content from Twitter

In the last decade, social networks like Facebook and Twitter increasingly became important parts of people´s lives. A rapidly growing number of users of these social networks create tremendous amounts of information, such as social text and multimedia feeds. In this context finding specific information and keeping it accessible becomes more and more a challenge. In parallel information filtering, topic modeling and quality has become subject of researchers in the field of social network analysis. Here we introduce Collact. Me, a conceptual framework to extract and classify microblogging content in an automated manner. Domain-specific data from social Twitter streams are collected from which topic models are created. The training data is used to build classifiers that allow computationally efficient multi-label classification. Results are presented and visualized using a novel dot-plot chart, which displays quantities of classified tweets of user-defined topics in a temporal fashion. We applied Collact. Me to selected bioinformatics topics. Our results show that our framework helps users to identify and interpret the level of attention of topics and to understand the relations between different topics as well as indications of emerging patterns. http://collact.me.